This is a LoRA for Taru's style, trained over his works across different periods, but mainly more current works.

The main benefit of using this over his artist tag@taaru \(taru\)) is no text or speech bubble is bundled with this LoRA, steam and similar effects also should not appear unprompted.

I might do another version if I have more time. This one might be a bit undercooked.

Recommended settings:

• cfg: 3

• sampler/scheduler: euler_a/beta

• steps: 30/12(with turbo lora)

• Optional recommended LoRAs: turbo v0.2: 0.8, rdbt v1 (or any anima base-v1.0 based versions): 0.5

• Trigger word: @taaru

• Optional trigger: @taaru \(taru\):0.5

• Recommended negative: white border, outside border, border

• Recommended meta tags: newest, recent, year 2025, year 2024, masterpiece, best quality, highres, absurdres

Training Info:

• dataset: 45 AI-processed img (repeat 2)

• tagger used: animetimm/caformer_b36.dbv4-full, Gemma 4 31b

• Epochs: 27

Known issues:

• pubic hair always present no matter prompt

  • Solution: Use pubic hair slider at strength -1

Side note: Using this LoRA together with @taaru \(taru\):0.5 seems to strengthen the style but text sometimes appear, if text does appear, try further lowering the weight of @taaru \(taru\). I also recommend doing a second sampler pass with hires fix/latent upscale(denoise: 0.5), it can fix most problems and imo make outputs better.

I will take this LoRA down on the Artist's request.